Repository navigation
Expand file tree
/
Copy pathtest_optimizer.py
More file actions
3257 lines (2829 loc) · 136 KB
/
Copy pathtest_optimizer.py
File metadata and controls
3257 lines (2829 loc) · 136 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
import unittest
from concurrent.futures import ProcessPoolExecutor, as_completed
from functools import partial
from unittest.mock import patch
import duckdb
from pandas.testing import assert_frame_equal
import sqlglot
from sqlglot import exp, optimizer, parse_one
from sqlglot.errors import ANSI_RESET, ANSI_UNDERLINE, OptimizeError, SchemaError
from sqlglot.optimizer.annotate_types import annotate_types
from sqlglot.optimizer.canonicalize_internal_names import canonicalize_internal_names
from sqlglot.optimizer.normalize import normalization_distance
from sqlglot.optimizer.qualify import qualify
from sqlglot.optimizer.scope import build_scope, traverse_scope, walk_in_scope
from sqlglot.schema import MappingSchema
from tests.helpers import (
TPCDS_SCHEMA,
TPCH_SCHEMA,
assert_logger_contains,
load_sql_fixture_pairs,
load_sql_fixtures,
string_to_bool,
)
def parse_and_optimize(func, sql, read_dialect, **kwargs):
return func(parse_one(sql, read=read_dialect), **kwargs)
def qualify_then_canonicalize(expression, **qualify_kwargs):
return canonicalize_internal_names(qualify(expression, **qualify_kwargs))
def qualify_columns(expression, validate_qualify_columns=True, **kwargs):
expression = optimizer.qualify.qualify(
expression,
infer_schema=True,
validate_qualify_columns=validate_qualify_columns,
identify=False,
**kwargs,
)
return expression
def pushdown_projections(expression, **kwargs):
expression = optimizer.qualify_tables.qualify_tables(expression)
expression = optimizer.qualify_columns.qualify_columns(expression, infer_schema=True, **kwargs)
expression = optimizer.pushdown_projections.pushdown_projections(expression, **kwargs)
return expression
def normalize(expression, **kwargs):
schema = kwargs.get("schema")
expression = optimizer.normalize.normalize(expression, dnf=False)
expression = annotate_types(expression, schema=schema)
return optimizer.simplify.simplify(expression)
def simplify(expression, **kwargs):
dialect = kwargs.get("dialect")
schema = kwargs.get("schema")
expression = annotate_types(expression, schema=schema)
return optimizer.simplify.simplify(
expression, constant_propagation=True, coalesce_simplification=True, dialect=dialect
)
def pushdown_ctes(expression, **kwargs):
optimizer.qualify_columns.pushdown_cte_alias_columns(build_scope(expression))
return expression
def annotate_functions(expression, **kwargs):
dialect = kwargs.get("dialect")
schema = kwargs.get("schema")
annotated = annotate_types(expression, dialect=dialect, schema=schema)
return annotated.expressions[0]
class TestOptimizer(unittest.TestCase):
maxDiff = None
@classmethod
def setUpClass(cls):
sqlglot.schema = MappingSchema()
cls.conn = duckdb.connect()
cls.conn.execute(
"""
CREATE TABLE x (a INT, b INT);
CREATE TABLE y (b INT, c INT);
CREATE TABLE z (b INT, c INT);
CREATE TABLE w (d TEXT, e TEXT);
INSERT INTO x VALUES (1, 1);
INSERT INTO x VALUES (2, 2);
INSERT INTO x VALUES (2, 2);
INSERT INTO x VALUES (3, 3);
INSERT INTO x VALUES (null, null);
INSERT INTO y VALUES (2, 2);
INSERT INTO y VALUES (2, 2);
INSERT INTO y VALUES (3, 3);
INSERT INTO y VALUES (4, 4);
INSERT INTO y VALUES (null, null);
INSERT INTO y VALUES (3, 3);
INSERT INTO y VALUES (3, 3);
INSERT INTO y VALUES (4, 4);
INSERT INTO y VALUES (5, 5);
INSERT INTO y VALUES (null, null);
INSERT INTO w VALUES ('a', 'b');
"""
)
def setUp(self):
self.schema = {
"x": {
"a": "INT",
"b": "INT",
},
"y": {
"b": "INT",
"c": "INT",
},
"z": {
"b": "INT",
"c": "INT",
},
"w": {
"d": "TEXT",
"e": "TEXT",
},
"temporal": {
"d": "DATE",
"t": "DATETIME",
},
"structs": {
"one": "STRUCT<a_1 INT, b_1 VARCHAR>",
"nested_0": "STRUCT<a_1 INT, nested_1 STRUCT<a_2 INT, nested_2 STRUCT<a_3 INT>>>",
"quoted": 'STRUCT<"foo bar" INT>',
},
"t_bool": {
"a": "BOOLEAN",
},
}
def check_file(
self,
file,
func,
pretty=False,
execute=False,
only=None,
**kwargs,
):
with ProcessPoolExecutor() as pool:
results = {}
for i, (meta, sql, expected) in enumerate(
load_sql_fixture_pairs(f"optimizer/{file}.sql"), start=1
):
title = meta.get("title") or f"{i}, {sql}"
if only and title != only:
continue
dialect = meta.get("dialect")
leave_tables_isolated = meta.get("leave_tables_isolated")
validate_qualify_columns = meta.get("validate_qualify_columns")
canonicalize_table_aliases = meta.get("canonicalize_table_aliases")
func_kwargs = kwargs.copy()
if leave_tables_isolated is not None:
func_kwargs["leave_tables_isolated"] = string_to_bool(leave_tables_isolated)
if validate_qualify_columns is not None:
func_kwargs["validate_qualify_columns"] = string_to_bool(
validate_qualify_columns
)
if dialect:
func_kwargs["dialect"] = dialect
if canonicalize_table_aliases is not None:
func_kwargs["canonicalize_table_aliases"] = string_to_bool(
canonicalize_table_aliases
)
future = pool.submit(parse_and_optimize, func, sql, dialect, **func_kwargs)
results[future] = (
sql,
title,
expected,
dialect,
execute if meta.get("execute") is None else False,
)
for future in as_completed(results):
sql, title, expected, dialect, execute = results[future]
with self.subTest(title):
optimized = future.result()
actual = optimized.sql(pretty=pretty, dialect=dialect)
self.assertEqual(
expected,
actual,
)
for expression in optimized.walk():
for arg_key, arg in expression.args.items():
if isinstance(arg, exp.Expr):
self.assertEqual(arg_key, arg.arg_key)
self.assertIs(arg.parent, expression)
if string_to_bool(execute):
with self.subTest(f"(execute) {title}"):
df1 = self.conn.execute(
sqlglot.transpile(sql, read=dialect, write="duckdb")[0]
).df()
df2 = self.conn.execute(optimized.sql(dialect="duckdb")).df()
assert_frame_equal(df1, df2)
@patch("sqlglot.generator.logger")
def test_optimize(self, logger):
self.assertEqual(optimizer.optimize("x = 1 + 1", identify=False).sql(), "x = 2")
schema = {
"x": {"a": "INT", "b": "INT"},
"y": {"b": "INT", "c": "INT"},
"z": {"a": "INT", "c": "INT"},
"u": {"f": "INT", "g": "INT", "h": "TEXT"},
}
self.check_file(
"optimizer",
optimizer.optimize,
infer_schema=True,
pretty=True,
execute=True,
schema=schema,
)
def test_isolate_table_selects(self):
self.check_file(
"isolate_table_selects",
optimizer.isolate_table_selects.isolate_table_selects,
schema=self.schema,
)
def test_qualify_tables(self):
tables = set()
optimizer.qualify.qualify(
parse_one("with foo AS (select * from bar) select * from foo join baz"),
qualify_columns=False,
on_qualify=lambda t: tables.add(t.name),
)
self.assertEqual(tables, {"bar", "baz"})
self.assertEqual(
optimizer.qualify.qualify(
parse_one("WITH tesT AS (SELECT * FROM t1) SELECT * FROM test", "bigquery"),
db="db",
catalog="catalog",
dialect="bigquery",
quote_identifiers=False,
).sql("bigquery"),
"WITH test AS (SELECT * FROM catalog.db.t1 AS t1) SELECT * FROM test AS test",
)
self.assertEqual(
optimizer.qualify_tables.qualify_tables(
parse_one(
"WITH cte AS (SELECT * FROM t) SELECT * FROM cte PIVOT(SUM(c) FOR v IN ('x', 'y'))"
),
db="db",
catalog="catalog",
).sql(),
"WITH cte AS (SELECT * FROM catalog.db.t AS t) SELECT * FROM cte AS cte PIVOT(SUM(c) FOR v IN ('x', 'y')) AS _0",
)
self.assertEqual(
optimizer.qualify_tables.qualify_tables(
parse_one(
"WITH cte AS (SELECT * FROM t) SELECT * FROM cte PIVOT(SUM(c) FOR v IN ('x', 'y')) AS pivot_alias"
),
db="db",
catalog="catalog",
).sql(),
"WITH cte AS (SELECT * FROM catalog.db.t AS t) SELECT * FROM cte AS cte PIVOT(SUM(c) FOR v IN ('x', 'y')) AS pivot_alias",
)
self.assertEqual(
optimizer.qualify_tables.qualify_tables(
parse_one("select a from b"), catalog="catalog"
).sql(),
"SELECT a FROM b AS b",
)
self.assertEqual(
optimizer.qualify_tables.qualify_tables(parse_one("select a from b"), db='"DB"').sql(),
'SELECT a FROM "DB".b AS b',
)
self.check_file(
"qualify_tables",
optimizer.qualify_tables.qualify_tables,
db="db",
catalog="c",
)
def test_qualify_tables_copies_typed_alias_columns(self):
expression = parse_one('SELECT * FROM JSON_TO_RECORDSET(z) AS y("rank" INT)')
original = expression.find(exp.Table).args["alias"].columns[0]
self.assertIsInstance(original, exp.ColumnDef)
optimizer.qualify_tables.qualify_tables(expression, canonicalize_table_aliases=True)
new = expression.find(exp.Table).args["alias"].columns[0]
self.assertIsInstance(new, exp.ColumnDef)
self.assertIsNot(original, new)
self.assertEqual(original.sql(), new.sql())
def test_normalize(self):
self.assertEqual(
optimizer.normalize.normalize(
parse_one("x AND (y OR z)"),
dnf=True,
).sql(),
"(x AND y) OR (x AND z)",
)
self.assertEqual(
optimizer.normalize.normalize(
parse_one("x AND (y OR z)"),
).sql(),
"x AND (y OR z)",
)
# Snowflake's BOOLXOR builds a Xor connector carrying a `round_input` arg, so unpacking
# via args.values() yields 3 values. The enclosing predicate isn't normalized, which forces
# _predicate_lengths to recurse into the Xor node.
self.assertEqual(
optimizer.normalize.normalize(
parse_one("(a AND b) OR BOOLXOR(x, y)", read="snowflake"),
).sql(dialect="snowflake"),
"((BOOLXOR(x, y)) OR a) AND ((BOOLXOR(x, y)) OR b)",
)
self.check_file("normalize", normalize, schema=self.schema)
@patch("sqlglot.generator.logger")
def test_qualify_columns(self, logger):
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"""
SELECT Teams.Name, count(*)
FROM raw.TeamMemberships as TeamMemberships
join raw.Teams
on Teams.Id = TeamMemberships.TeamId
GROUP BY 1
""",
read="bigquery",
),
schema={
"raw": {
"TeamMemberships": {
"Id": "INTEGER",
"UserId": "INTEGER",
"TeamId": "INTEGER",
},
"Teams": {
"Id": "INTEGER",
"Name": "STRING",
},
}
},
dialect="bigquery",
).sql(dialect="bigquery"),
"SELECT `teams`.`name` AS `name`, count(*) AS `_col_1` FROM `raw`.`TeamMemberships` AS `teammemberships` JOIN `raw`.`Teams` AS `teams` ON `teams`.`id` = `teammemberships`.`teamid` GROUP BY `teams`.`name`",
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"SELECT `my_db.my_table`.`my_column` FROM `my_db.my_table`",
read="bigquery",
),
dialect="bigquery",
).sql(dialect="bigquery"),
"SELECT `my_table`.`my_column` AS `my_column` FROM `my_db.my_table` AS `my_table`",
)
# Alias pushdown reaches VALUES set operation operands, since the parser wraps them in
# selects; the union's output names come from its left branch, so this is required for
# the CTE's column list to propagate correctly
self.assertEqual(
optimizer.qualify.qualify(
parse_one("WITH t(a) AS (VALUES (1) UNION ALL SELECT 2) SELECT t.a FROM t")
).sql(),
'WITH "t" AS (SELECT "_values"."_col_0" AS "a" FROM (VALUES (1)) AS "_values"("_col_0") UNION ALL SELECT 2 AS "a") SELECT "t"."a" AS "a" FROM "t" AS "t"',
)
# In the recursive branch of a recursive CTE, an unqualified column that matches one of
# the CTE's columns must not be expanded to the projection alias it shadows, e.g. here
# `n < 5` refers to t.n, not to `n + 1` (tested in Postgres and DuckDB)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"WITH RECURSIVE t(n) AS (SELECT 1 AS n UNION ALL SELECT n + 1 AS n FROM t WHERE n < 5) SELECT * FROM t"
)
).sql(),
'WITH RECURSIVE "t"("n") AS (SELECT 1 AS "n" UNION ALL SELECT "t"."n" + 1 AS "n" FROM "t" AS "t" WHERE "t"."n" < 5) SELECT "t"."n" AS "n" FROM "t" AS "t"',
)
# Same, but with a parenthesized recursive part: its scope is the inner query, so the
# union's right arm must be unnested for the check above to kick in
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"WITH RECURSIVE t(n) AS (SELECT 1 AS n UNION ALL (SELECT n + 1 AS n FROM t WHERE n < 5)) SELECT * FROM t"
)
).sql(),
'WITH RECURSIVE "t"("n") AS (SELECT 1 AS "n" UNION ALL (SELECT "t"."n" + 1 AS "n" FROM "t" AS "t" WHERE "t"."n" < 5)) SELECT "t"."n" AS "n" FROM "t" AS "t"',
)
# Programmatic set operations with VALUES operands must not crash the Query-typed
# left/right accessors in compiled builds
with self.assertRaises(OptimizeError):
optimizer.qualify.qualify(
exp.select("*").from_(
exp.Union(
this=exp.select("a").from_("x"),
expression=parse_one("VALUES (2)"),
distinct=False,
side="LEFT",
).subquery("s")
)
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"SELECT pos, val FROM t CROSS JOIN LATERAL (SELECT pos - 1 AS pos, val FROM UNNEST(t.arr) WITH ORDINALITY AS _t0(val, pos))",
read="duckdb",
),
schema={"t": {"arr": "ARRAY<VARCHAR>"}},
dialect="duckdb",
).sql(dialect="duckdb"),
'SELECT "_0"."pos" AS "pos", "_0"."val" AS "val" FROM "t" AS "t" CROSS JOIN LATERAL (SELECT "_t0"."pos" - 1 AS "pos", "_t0"."val" AS "val" FROM UNNEST("t"."arr") WITH ORDINALITY AS "_t0"("val", pos)) AS "_0"',
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"SELECT * FROM t CROSS JOIN LATERAL (SELECT 1 AS x, 2 AS y) AS foo",
read="duckdb",
),
schema={"t": {"k": "INT"}},
dialect="duckdb",
).sql(dialect="duckdb"),
'SELECT "t"."k" AS "k", "foo"."x" AS "x", "foo"."y" AS "y" FROM "t" AS "t" CROSS JOIN LATERAL (SELECT 1 AS "x", 2 AS "y") AS "foo"',
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"SELECT c, d FROM t CROSS JOIN LATERAL (SELECT 1 AS a, 2 AS b) AS x(c, d)",
read="duckdb",
),
schema={"t": {"k": "INT"}},
dialect="duckdb",
).sql(dialect="duckdb"),
'SELECT "x"."c" AS "c", "x"."d" AS "d" FROM "t" AS "t" CROSS JOIN LATERAL (SELECT 1 AS "a", 2 AS "b") AS "x"("c", "d")',
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"SELECT * FROM t CROSS JOIN LATERAL (SELECT 1 AS a, 2 AS b) AS x(c)",
read="duckdb",
),
schema={"t": {"k": "INT"}},
dialect="duckdb",
).sql(dialect="duckdb"),
'SELECT "t"."k" AS "k", "x"."c" AS "c", "x"."b" AS "b" FROM "t" AS "t" CROSS JOIN LATERAL (SELECT 1 AS "a", 2 AS "b") AS "x"("c")',
)
self.assertEqual(
optimizer.qualify_columns.qualify_columns(
parse_one(
"WITH RECURSIVE t AS (SELECT 1 AS x UNION ALL SELECT x + 1 FROM t AS child WHERE x < 10) SELECT * FROM t"
),
schema={},
infer_schema=False,
).sql(),
"WITH RECURSIVE t AS (SELECT 1 AS x UNION ALL SELECT child.x + 1 AS _col_0 FROM t AS child WHERE child.x < 10) SELECT t.x AS x FROM t",
)
self.assertEqual(
optimizer.qualify_columns.qualify_columns(
parse_one("WITH x AS (SELECT a FROM db.y) SELECT * FROM db.x"),
schema={"db": {"x": {"z": "int"}, "y": {"a": "int"}}},
expand_stars=False,
).sql(),
"WITH x AS (SELECT y.a AS a FROM db.y) SELECT * FROM db.x",
)
self.assertEqual(
optimizer.qualify_columns.qualify_columns(
parse_one("WITH x AS (SELECT a FROM db.y) SELECT z FROM db.x"),
schema={"db": {"x": {"z": "int"}, "y": {"a": "int"}}},
infer_schema=False,
).sql(),
"WITH x AS (SELECT y.a AS a FROM db.y) SELECT x.z AS z FROM db.x",
)
self.assertEqual(
optimizer.qualify_columns.qualify_columns(
parse_one("select y from x"),
schema={},
infer_schema=False,
).sql(),
"SELECT y AS y FROM x",
)
# Aliases derived from quoted projections keep their exact spelling, even for
# dialects that fold quoted identifiers (e.g. Trino), so that running this rule
# without `normalize_identifiers` doesn't desync the alias from its source
self.assertEqual(
optimizer.qualify_columns.qualify_columns(
parse_one('SELECT "C1" FROM t', read="trino"),
schema={},
infer_schema=False,
dialect="trino",
).sql(dialect="trino"),
'SELECT "C1" AS "C1" FROM t',
)
self.assertEqual(
optimizer.qualify_columns.qualify_columns(
parse_one('SELECT ("C1") FROM t', read="trino"),
schema={},
infer_schema=False,
dialect="trino",
).sql(dialect="trino"),
'SELECT ("C1") AS "C1" FROM t',
)
self.assertEqual(
optimizer.qualify_columns.qualify_columns(
parse_one('SELECT t.s."Ff" FROM t', read="duckdb"),
schema={"t": {"s": 'STRUCT("Ff" INT)'}},
dialect="duckdb",
).sql(dialect="duckdb"),
'SELECT t.s."Ff" AS "Ff" FROM t',
)
# Same for unquoted identifiers: folding only the derived alias would leave the
# outer scope unable to resolve (and qualify) references to the inner output
self.assertEqual(
optimizer.qualify_columns.qualify_columns(
parse_one("SELECT Cc FROM (SELECT Cc FROM t) AS s", read="snowflake"),
schema={},
infer_schema=False,
dialect="snowflake",
).sql(dialect="snowflake"),
"SELECT s.Cc AS Cc FROM (SELECT Cc AS Cc FROM t) AS s",
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"WITH X AS (SELECT Y.A FROM DB.y CROSS JOIN a.b.INFORMATION_SCHEMA.COLUMNS) SELECT `A` FROM X",
read="bigquery",
),
dialect="bigquery",
).sql(),
'WITH "x" AS (SELECT "y"."a" AS "a" FROM "DB"."y" AS "y" CROSS JOIN "a"."b"."INFORMATION_SCHEMA.COLUMNS" AS "columns") SELECT "x"."a" AS "a" FROM "x" AS "x"',
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"CREATE FUNCTION udfs.`myTest`(`x` FLOAT64) AS (1)",
read="bigquery",
),
dialect="bigquery",
).sql(dialect="bigquery"),
"CREATE FUNCTION `udfs`.`myTest`(`x` FLOAT64) AS (1)",
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one("SELECT `bar_bazfoo_$id` FROM test", read="spark"),
schema={"test": {"bar_bazFoo_$id": "BIGINT"}},
dialect="spark",
).sql(dialect="spark"),
"SELECT `test`.`bar_bazfoo_$id` AS `bar_bazfoo_$id` FROM `test` AS `test`",
)
qualified = optimizer.qualify.qualify(
parse_one("WITH t AS (SELECT 1 AS c) (SELECT c FROM t)")
)
self.assertIs(qualified.selects[0].parent, qualified)
self.assertEqual(
qualified.sql(),
'WITH "t" AS (SELECT 1 AS "c") SELECT "t"."c" AS "c" FROM "t" AS "t"',
)
# can't coalesce USING columns because they don't exist in every already-joined table
self.assertEqual(
optimizer.qualify_columns.qualify_columns(
parse_one(
"SELECT id, dt, v FROM (SELECT t1.id, t1.dt, sum(coalesce(t2.v, 0)) AS v FROM t1 AS t1 LEFT JOIN lkp AS lkp USING (id) LEFT JOIN t2 AS t2 USING (other_id, dt, common) WHERE t1.id > 10 GROUP BY 1, 2) AS `_0`",
dialect="bigquery",
),
schema=MappingSchema(
schema={
"t1": {"id": "int64", "dt": "date", "common": "int64"},
"lkp": {"id": "int64", "other_id": "int64", "common": "int64"},
"t2": {"other_id": "int64", "dt": "date", "v": "int64", "common": "int64"},
},
dialect="bigquery",
),
).sql(dialect="bigquery"),
"SELECT `_0`.id AS id, `_0`.dt AS dt, `_0`.v AS v FROM (SELECT t1.id AS id, t1.dt AS dt, sum(coalesce(t2.v, 0)) AS v FROM t1 AS t1 LEFT JOIN lkp AS lkp ON t1.id = lkp.id LEFT JOIN t2 AS t2 ON lkp.other_id = t2.other_id AND t1.dt = t2.dt AND COALESCE(t1.common, lkp.common) = t2.common WHERE t1.id > 10 GROUP BY t1.id, t1.dt) AS `_0`",
)
# Detection of correlation where columns are referenced in derived tables nested within subqueries
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"SELECT a.g FROM a WHERE a.e < (SELECT MAX(u) FROM (SELECT SUM(c.b) AS u FROM c WHERE c.d = f GROUP BY c.e) w)"
),
schema={
"a": {"g": "INT", "e": "INT", "f": "INT"},
"c": {"d": "INT", "e": "INT", "b": "INT"},
},
quote_identifiers=False,
).sql(),
"SELECT a.g AS g FROM a AS a WHERE a.e < (SELECT MAX(w.u) AS _col_0 FROM (SELECT SUM(c.b) AS u FROM c AS c WHERE c.d = a.f GROUP BY c.e) AS w)",
)
# Detection of correlation where columns are referenced in derived tables nested within lateral joins
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"SELECT u.user_id, l.log_date FROM users AS u CROSS JOIN LATERAL (SELECT l1.log_date FROM (SELECT l.log_date FROM logs AS l WHERE l.user_id = u.user_id AND l.log_date <= 100 ORDER BY l.log_date LIMIT 1) AS l1) AS l",
dialect="postgres",
),
schema={
"users": {"user_id": "text", "log_date": "date"},
"logs": {"user_id": "text", "log_date": "date"},
},
quote_identifiers=False,
).sql("postgres"),
"SELECT u.user_id AS user_id, l.log_date AS log_date FROM users AS u CROSS JOIN LATERAL (SELECT l1.log_date AS log_date FROM (SELECT l.log_date AS log_date FROM logs AS l WHERE l.user_id = u.user_id AND l.log_date <= 100 ORDER BY l.log_date LIMIT 1) AS l1) AS l",
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"SELECT A.b_id FROM A JOIN B ON A.b_id=B.b_id JOIN C USING(c_id)",
dialect="postgres",
),
schema={
"A": {"b_id": "int"},
"B": {"b_id": "int", "c_id": "int"},
"C": {"c_id": "int"},
},
quote_identifiers=False,
).sql("postgres"),
"SELECT a.b_id AS b_id FROM a AS a JOIN b AS b ON a.b_id = b.b_id JOIN c AS c ON b.c_id = c.c_id",
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"SELECT A.b_id FROM A JOIN B ON A.b_id=B.b_id JOIN C ON B.b_id = C.b_id JOIN D USING(d_id)",
dialect="postgres",
),
schema={
"A": {"b_id": "int"},
"B": {"b_id": "int", "d_id": "int"},
"C": {"b_id": "int"},
"D": {"d_id": "int"},
},
quote_identifiers=False,
).sql("postgres"),
"SELECT a.b_id AS b_id FROM a AS a JOIN b AS b ON a.b_id = b.b_id JOIN c AS c ON b.b_id = c.b_id JOIN d AS d ON b.d_id = d.d_id",
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"""
SELECT
(SELECT SUM(c.amount)
FROM UNNEST(credits) AS c
WHERE type != 'promotion') as total
FROM billing
""",
read="bigquery",
),
schema={"billing": {"credits": "ARRAY<STRUCT<amount FLOAT64, type STRING>>"}},
dialect="bigquery",
).sql(dialect="bigquery"),
"SELECT (SELECT SUM(`c`.`amount`) AS `_col_0` FROM UNNEST(`billing`.`credits`) AS `c` WHERE `type` <> 'promotion') AS `total` FROM `billing` AS `billing`",
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"""
WITH cte AS (SELECT * FROM base_table)
SELECT
(SELECT SUM(item.price)
FROM UNNEST(items) AS item
WHERE category = 'electronics') as electronics_total
FROM cte
""",
read="bigquery",
),
schema={
"base_table": {
"id": "INT64",
"items": "ARRAY<STRUCT<price FLOAT64, category STRING>>",
}
},
dialect="bigquery",
).sql(dialect="bigquery"),
"WITH `cte` AS (SELECT `base_table`.`id` AS `id`, `base_table`.`items` AS `items` FROM `base_table` AS `base_table`) SELECT (SELECT SUM(`item`.`price`) AS `_col_0` FROM UNNEST(`cte`.`items`) AS `item` WHERE `category` = 'electronics') AS `electronics_total` FROM `cte` AS `cte`",
)
self.check_file(
"qualify_columns",
qualify_columns,
execute=True,
schema=self.schema,
)
self.check_file("qualify_columns_ddl", qualify_columns, schema=self.schema)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"""
SELECT
(
SELECT
col_st.value
FROM UNNEST(col_st) AS col_st
) AS vcol1
FROM t AS b
""",
read="bigquery",
),
schema={
"t": {
"col_st": "ARRAY<STRUCT<key STRING, value INT>>",
}
},
dialect="bigquery",
).sql(dialect="bigquery"),
"SELECT (SELECT `col_st`.`value` AS `value` FROM UNNEST(`b`.`col_st`) AS `col_st`) AS `vcol1` FROM `t` AS `b`",
)
# Schema-qualified table joined twice (once unaliased, once aliased) should resolve correctly
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"SELECT 1 FROM dbo.a JOIN dbo.b ON dbo.b.id = dbo.a.id JOIN dbo.b AS x ON x.id = dbo.a.id"
),
).sql(),
'SELECT 1 AS "1" FROM "dbo"."a" AS "a" JOIN "dbo"."b" AS "b" ON "b"."id" = "a"."id" JOIN "dbo"."b" AS "x" ON "x"."id" = "a"."id"',
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one("SELECT * FROM t"),
schema={"t": {"end": "text"}},
quote_identifiers=False,
).sql(),
"SELECT t.end AS end FROM t AS t",
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"WITH produce AS (SELECT 'Kale' AS product, 51 AS q1, 23 AS q2) "
"SELECT * FROM produce UNPIVOT(sales FOR quarter IN (q1, q2))",
dialect="bigquery",
),
dialect="bigquery",
).sql(dialect="bigquery"),
"WITH `produce` AS (SELECT 'Kale' AS `product`, 51 AS `q1`, 23 AS `q2`) "
"SELECT `produce`.`product` AS `product`, `produce`.`quarter` AS `quarter`, "
"`produce`.`sales` AS `sales` FROM `produce` AS `produce` "
"UNPIVOT(`sales` FOR `quarter` IN (`produce`.`q1`, `produce`.`q2`)) AS `produce`",
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"WITH cte AS (SELECT 1 AS a, 2 AS b, 3 AS c) "
"SELECT u.val, u.name FROM cte UNPIVOT(val FOR name IN (a, b, c)) AS u"
),
).sql(),
'WITH "cte" AS (SELECT 1 AS "a", 2 AS "b", 3 AS "c") '
'SELECT "u"."val" AS "val", "u"."name" AS "name" FROM "cte" AS "cte" '
'UNPIVOT("val" FOR "name" IN ("cte"."a", "cte"."b", "cte"."c")) AS "u"',
)
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"WITH produce AS (SELECT 'Kale' AS product, 51 AS q1, 23 AS q2, 45 AS q3, 3 AS q4) "
"SELECT * FROM produce UNPIVOT((first_half, second_half) FOR semesters "
"IN ((q1, q2) AS 'h1', (q3, q4) AS 'h2'))",
dialect="bigquery",
),
dialect="bigquery",
).sql(dialect="bigquery"),
"WITH `produce` AS (SELECT 'Kale' AS `product`, 51 AS `q1`, 23 AS `q2`, 45 AS `q3`, 3 AS `q4`) "
"SELECT `produce`.`product` AS `product`, `produce`.`semesters` AS `semesters`, "
"`produce`.`first_half` AS `first_half`, `produce`.`second_half` AS `second_half` "
"FROM `produce` AS `produce` UNPIVOT((`first_half`, `second_half`) FOR `semesters` "
"IN ((`produce`.`q1`, `produce`.`q2`) AS 'h1', (`produce`.`q3`, `produce`.`q4`) AS 'h2')) AS `produce`",
)
def test_unnest_type_trace_is_memoized(self):
"""Tracing an UNNEST's element type must not re-walk shared parts of the scope graph.
qualify resolves an unnested column's type by walking it back to a base table
(Resolver._get_column_type_from_scope). Each order_attrs_k joins order_attrs_{k-1}
back to the shared `orders` CTE, which does not hold `attrs`, so the walk cannot
short-circuit and revisits it through every path. Without memoization the trace is
called ~93k times at 12 levels; with it, ~90.
"""
from sqlglot.optimizer import resolver as resolver_module
n_levels = 12
ctes = [
"order_attrs0 AS (SELECT id, [STRUCT('color' AS key, 'red' AS value)] AS attrs FROM orders)"
]
for k in range(1, n_levels):
prev = f"order_attrs{k - 1}"
ctes.append(
f"order_attrs{k} AS (SELECT a.id AS id, "
f"array_concat(a.attrs, [STRUCT('size' AS key, 'large' AS value)]) AS attrs "
f"FROM {prev} AS a JOIN orders AS o ON a.id = o.id)"
)
ctes.append(
f"non_null_attrs AS (SELECT ARRAY(SELECT x FROM UNNEST(attrs) AS x "
f"WHERE NOT x.value IS NULL) AS attrs FROM order_attrs{n_levels - 1})"
)
sql = "WITH " + ", ".join(ctes) + " SELECT * FROM non_null_attrs"
original = resolver_module.Resolver._get_column_type_from_scope
with patch.object(
resolver_module.Resolver,
"_get_column_type_from_scope",
autospec=True,
side_effect=original,
) as traced:
optimizer.qualify.qualify(
parse_one(sql, dialect="bigquery"),
schema={"orders": {"id": "INT64"}},
dialect="bigquery",
)
self.assertLess(
traced.call_count,
1000,
f"got {traced.call_count} trace calls -- memoization may be broken",
)
def test_validate_columns(self):
with self.assertRaisesRegex(
OptimizeError, "Column 'foo' could not be resolved. Line: 1, Col: 10"
):
optimizer.qualify.qualify(
parse_one("select foo from x"),
schema={"foo": {"y": "int"}},
)
# Test ambiguous columns error with PIVOT (which skips "could not be resolved" check)
with self.assertRaisesRegex(OptimizeError, "Ambiguous column 'a'"):
expression = parse_one(
"SELECT * FROM (SELECT a, b, c FROM x) PIVOT (SUM(b) FOR c IN ('x', 'y'))"
)
qualified = optimizer.qualify_columns.qualify_columns(
expression, schema={"x": {"a": "int", "b": "int", "c": "str"}}
)
optimizer.qualify_columns.validate_qualify_columns(qualified)
schema = {"my_table": {"items": "ARRAY<STRUCT<name STRING, age INT>>"}}
expression = annotate_types(
optimizer.qualify.qualify(
parse_one(
"SELECT ci.name, ci.age FROM my_table LATERAL VIEW EXPLODE(items) ci AS ci",
read="spark",
),
schema=schema,
dialect="spark",
),
schema=schema,
dialect="spark",
)
self.assertEqual(
expression.sql(dialect="spark"),
"SELECT `ci`.`name` AS `name`, `ci`.`age` AS `age` FROM `my_table` AS `my_table` LATERAL VIEW EXPLODE(`my_table`.`items`) ci AS `ci`",
)
self.assertEqual(expression.selects[0].type, exp.DataType.build("STRING", dialect="spark"))
self.assertEqual(expression.selects[1].type, exp.DataType.build("INT", dialect="spark"))
schema = {"my_table": {"items": "ARRAY<STRUCT<amount FLOAT, type STRING>>"}}
expression = annotate_types(
optimizer.qualify.qualify(
parse_one(
"SELECT (SELECT SUM(ci.amount) FROM my_table LATERAL VIEW EXPLODE(items) ci AS ci WHERE ci.type = 'promotion') AS total FROM my_table",
read="spark",
),
schema=schema,
dialect="spark",
),
schema=schema,
dialect="spark",
)
self.assertEqual(
expression.sql(dialect="spark"),
"SELECT (SELECT SUM(`ci`.`amount`) AS `_col_0` FROM `my_table` AS `my_table` LATERAL VIEW EXPLODE(`my_table`.`items`) ci AS `ci` WHERE `ci`.`type` = 'promotion') AS `total` FROM `my_table` AS `my_table`",
)
self.assertEqual(expression.selects[0].type, exp.DataType.build("DOUBLE", dialect="spark"))
# An unqualified struct field is disambiguated through the lateral's extended columns
schema = {"my_table": {"items": "ARRAY<STRUCT<name STRING, age INT>>"}}
self.assertEqual(
optimizer.qualify.qualify(
parse_one(
"SELECT name FROM my_table LATERAL VIEW EXPLODE(items) ci AS ci",
read="spark",
),
schema=schema,
dialect="spark",
).sql(dialect="spark"),
"SELECT `ci`.`name` AS `name` FROM `my_table` AS `my_table` LATERAL VIEW EXPLODE(`my_table`.`items`) ci AS `ci`",
)
# Resolving an unqualified lateral column whose table is missing from the schema must
# raise instead of recursing infinitely
with self.assertRaisesRegex(OptimizeError, "Column 'ITEMS' could not be resolved"):
optimizer.qualify.qualify(
parse_one(
"SELECT f.value AS v FROM my_db.raw.events, LATERAL FLATTEN(items) AS f",
read="snowflake",
),
schema={"my_db": {"other": {"some_view": {"v": "VARIANT"}}}},
dialect="snowflake",
)
def test_qualify_columns_struct_star_expansion_types(self):
# Struct star expansion annotates a scope and then replaces its projections; the
# stale annotation caches must not prevent the new projections from being typed
qualified = optimizer.qualify.qualify(
parse_one(
"WITH cte AS (SELECT t.s.* FROM t UNION ALL SELECT t.s.* FROM t) SELECT cte.x FROM cte",
read="bigquery",
),
schema={"t": {"s": "STRUCT<x INT64, y STRING>"}},
dialect="bigquery",
)
union = qualified.find(exp.Union)
assert union
self.assertEqual(
[s.type.sql("bigquery") if s.type else None for s in union.selects],
["INT64", "STRING"],
)
self.assertEqual(
[s.type.sql("bigquery") if s.type else None for s in union.expression.selects],
["INT64", "STRING"],
)
self.assertEqual(qualified.selects[0].type.sql("bigquery"), "INT64")
def test_qualify_columns__with_invisible(self):
schema = MappingSchema(self.schema, {"x": {"a"}, "y": {"b"}, "z": {"b"}})
self.check_file("qualify_columns__with_invisible", qualify_columns, schema=schema)
def test_pushdown_cte_alias_columns(self):
self.check_file(
"pushdown_cte_alias_columns",
pushdown_ctes,
)
def test_qualify_columns__invalid(self):